Re: Google Summer of Code 2009: GIT

From: Rogan Dawes <lists@dawes.za.net>
Date: Mar 11, 2009, 15:27 UTC
Message-ID: <49B7D84B.6080501@dawes.za.net>
In-Reply-To: <ab9fa62a0903110713k2a21cefbj1e7cd3c126aca8f9@mail.gmail.com>

saurabh gupta wrote:

Show 17 quoted lines

>>> However, I think in merging and notifying about the conflicts in the xml
>>> files, other things can also be put forward. Like the GUI will show the
>>> number of tags differing and what are the new tags added and even if any
>>> tag is renamed with the content unchanged. If possible, how about
>>> showing a tree like structure (just like DOM model) to compare (or diff)
>>> the two xml files.
>>
>> This is a little bit too low-level for my liking.  Taking the OpenOffice
>> example again, the GUI should not expose XML at all...
> 
> hmmmm.....I think I get your point somewhat. Let me do some research
> over the formats and the background formats in which tools like
> OpenOffice store the data in xml files. May be for docbooks by
> OpenOffice, the best thing would be to give the *diff* output in terms
> of lines.
> I would also appreciate to know what you think and would like to see
> the output in such case.

I think that the implementation may make use of features inherent in the file format where possible. e.g. I suspect that OpenOffice has the ability to show "Tracked changes", and then allow the user to view the changes using the actual OpenOffice implementation.

I suspect that that will get a lot more difficult with e.g. conflicts and merges, because I doubt that OOo has the ability to show changes from multiple versions.

But I have to agree with Dscho, that the output would have to depend on the file type (OOo document), not just the data structure (e.g. XML) inside the file.

A regular XML file diff could choose to ignore/collapse whitespace (pretty printing) when doing the comparison, to show things like moving a branch further down the tree.

e.g.

<i>text</i>

vs

<b><i>text</i></b>

vs

<b>
  <i>text</i>
</b>

For plain XML, a textual diff might choose to show it with each element un-indented, and a standard text diff output:

+ <b>
  <i>
  text
  </i>
+ </b>

while a GUI diff might show the new element highlighted in a tree:

#green#<b>#/green#
  <i>
   text

I think that where reasonable that you should aim to have a text-only version that could be wrapped by a GUI. Obviously, this would be meaningless when diffing a JPG, for instance.

Ok, that was a bit rambling. I hope it helped more than it confused.

Rogan